Have we entered the age of NoOps infrastructures? Hardly. Old-style system administrators may be disappearing in the face of automation and cloud computing, but operations have become more significant than ever. As this O’Reilly Radar Report explains, we’re moving into a more complex arrangement known as "DevOps."
Mike Loukides, O’Reilly’s VP of Content Strategy, provides an incisive look into this new world of operations, where IT specialists are becoming part of the development team. In an environment with thousands of servers, these specialists now write the code that maintains the infrastructure. Even applications that run in the cloud have to be resilient and fault tolerant, need to be monitored, and must adjust to huge swings in load. That was underscored by Amazon’s EBS outage last year.
From the discussions at O’Reilly’s Velocity Conference, it’s evident that many operations specialists are quickly adapting to the DevOps reality. But as a whole, the industry has just scratched the surface. This report tells you why.
Mike Loukides is an editor for O'Reilly & Associates. He is the author of System Performance Tuning and UNIX for FORTRAN Programmers. Mike's interests are system administration, networking, programming languages, and computer architecture. His academic background includes degrees in electrical engineering (B.S.) and English literature (Ph.D.).
Mike Loukides takes on a hot topic in his latest treaty, What is DevOps? For those unfamiliar to the debate, several IT professionals have weighed in on this meaty subject, with a starting salvo from Netflix's Adrian Cockcroft followed by a counterpoint from Etsy's John Allspaw.
The crux of the debate revolves around the definition of DevOps. For those of us with one foot in the development community, and one in the operations/system administration side, it can seem like a tempest in a teapot, but figuring out how best to orient development versus operations is important because of a crucial factor: cost.
Loukides walks through the history of development and operations, and ultimately comes down mostly on Allspaw's side, opining that regardless of what you call DevOps (and he feels that NoOps is a poor choice of words) that it will continue to thrive and flourish as the field matures, which can only be a good thing.
Largely, I agree with with Loukides that having a robust, cloud-centered framework with repeatable processes and "infrastructure as code" is definitely the direction organizations are heading. At the same time, only a few very large organizations dealing with incredibly large datasets that need to be highly available are pushing boldly into this brand new world. That's not due to a lack of platforms, of course, and the recent announcement of Microsoft to provide virtual machines as part of their cloud platform, Azure, illustrates the current tension in the community at large. Although a Netflix or an Etsy may need thousands of nodes working together to provide a highly distributed, highly available service, most companies aren't that sophisticated. As such, they may just want to virtualize a few (hundred?) machines and worry about porting their workloads at a later point.
Here at New Signature we've been excited about Azure's ability to transform organizations with good developers on staff into amazingly dynamic organizations. For those groups that are unfamiliar with the tenets of Platform-As-A-Service (PaaS), though, the ability to spin up virtual machines in a cloud infrastructure may provide the necessary rationale to dip their toes in the water. For these companies, having a great Infrastructure-As-A-Services (IaaS) is even more critical because it can help bridge the gap. Not having to worry about physical network switches or physical boxes is a giant leap forward, even if applications haven't been rearchitected to be fully distributed and highly available.
Thus, for 90% of the organizations that are out there, DevOps is a critical goal, rather than an easily attainable action item.
Bottom Line Yes, I would recommend this to a friend